Taiwan released AI SARS-CoV-2 classifier with NHIA and CDC

Abstract

Taiwan AI Labs is working with the Executive Yuan, the Centers for Disease Control (CDC), and the National Health Insurance Administration (NHIA) to deploy an automated SARS-CoV-2 screening model. Hospitals in Taiwan file Chest X-Ray (CXR) images through the service of NHIA. The model will trigger alerts to notify CDC when high-confident positive cases identified.

Background

Since the outbreak of SARS-CoV-2, there are limited confirmed cases in Taiwan and only few doctors experienced in diagnosing SARS-CoV-2. It usually takes a couple days to reach a diagnosis result by doctors with cases showing no supplemental symptoms. To tackle this problem, NTUH healthcare experts suggested a centralized AI-assisted CXR screening solution. Instead of CT and other modalities, CXR is recommended because of its efficiency and wider availability.

Automated CXR data collection from NHIA

Taiwan has 478 hospitals in total for its 24 million population. The NHIA covers 99.7% of the population, maintaining a centralized system to collect medical data from all insured hospitals with high-standard security and privacy. Since the outbreak of SARS-CoV-2 in December 2019, NHIA has received more than 5000 CXR images.

Deployed AI-based SARS-CoV-2 screening model as an open API

Collaborated with hospitals in Taiwan, the SARS-CoV-2 CXR screening model developed by Taiwan AI Labs reached high sensitivity and specificity for pneumonia and SARS-CoV-2 recognition. The model is now an open API for hospitals around the world.

Integrated with CDC for potential candidates screening

The Executive Yuan and Taiwan AI Labs took a step further to integrate this pneumonia and SARS-CoV-2 screening model into the internal system of NHIA. Once a CXR uploaded to the server, protected by the same NHIA security standard, the AI-based SARS-CoV-2 screening model will be triggered. The result of this system will be sent to the CDC immediately for further conduct whenever a high-confidence positive case of SARS-CoV-2 is detected.

The result of SARS-CoV-2 screening model simulated on CXR in NHIA taken in the last 3 months (January 2020 to March 2020)

  • Case 24 was detected 17 days earlier before PCR confirmed (on January 31, 2020).
  • Case 268 was detected 6 days earlier before PCR confirmed (on March 20, 2020).
  • There were 27 out of 109 cases detected at least 2 days earlier than the PCR confirmed dates (Case 24, 226, 227, 235, 256, 268, 271, 272, 273, 279, 284, 287, 289, 290, 292, 308, 310, 313, 316, 317, 335, 349, 353, 355, 361, 363, and 379). Among these cases, Case 24, 268, 279, and 292 were detected 5 days earlier.

To test the classifier, please access via HERE (Username: guest / Password: guest)